Autonomous technologies: an uncertain and evolving risk landscape

Propelled by massive investments and impressive technological advances, autonomous vehicles are quickly becoming more numerous and capable, and appear in a wider variety of settings.

Today, mobile autonomy applications include semi-autonomous to fully self-driving cars, taxis, buses and trucks operating on public roads; shuttle vans moving people in closed environments such as airports and campus settings; automated forklifts ferrying goods around distribution centres and manufacturing operations; mini pods delivering groceries within defined territories; autonomous farm machines planting and harvesting crops; and the list goes on.

Replicating human abilities

Autonomous vehicles are exceptionally complex systems with many advanced components and sub-components. Regardless of where and how they’re used, these machines have to be capable of knowing: Where am I? What’s around me? What do I do next? They also have to answer those questions instantly, consistently, precisely and, in many cases, in varied and dynamic environments.

In other words, mobile autonomy requires replicating human perception, cognition, and our ability to take precise and appropriate action. These qualities are being reliably duplicated in self-driving vehicles and robots with:

Hardware including cameras, radar and Lidar sensors that record a vehicle’s location and surroundings as well as the prevailing environmental conditions

Software code and algorithms that process the various data stream

Automated controls for steering, acceleration and braking.

Installing these human abilities inside a machine thus requires a lot of sophisticated hardware and software working together seamlessly and unfailingly. In fact, automobiles today with limited self-driving capabilities already contain up to one hundred electronic-control units and ten kilometres of cabling, and these vehicles rely on around a million lines of code to operate. Moreover, increasingly popular in-car entertainment systems, while separate from vehicle performance, represent a gateway for product recalls resulting from cyber-security risks.

An evolving risk landscape

As this sector continues to develop, accidents are projected to decline markedly. In the U.S., for instance, distracted driving claimed 3,166 lives in 2017. Likewise, a Dutch study estimated that eliminating mobile phone use while driving would prevent about 600 road deaths annually. Machines, on the other hand, are never drowsy or distracted. Take the steering wheel away from people, and the number of accidents should plummet.

While the safety benefits of mobile autonomy are likely to be significant, the risk management issues and implications for companies in this ecosystem, including vehicle manufacturers, fleet owners/operators, component manufacturers and software developers, are still coming into focus.

Some observers predict, for instance, that product recalls, a historical challenge for the automotive industry, could take on even greater prominence as autonomous vehicles are deployed for different purposes. According to recent research conducted by a team that included one of my colleagues, “… product recall events are increasing in general, but recall events associated with (complex autonomous vehicle) technology form an increasingly large percentage of these recall events”.

In addition to quantifying the increased frequency of product recalls involving autonomous technologies, the researchers also concluded: “At the same time, financial risks resulting from extensive product recall events can severely affect vehicle manufacturers and their suppliers, exposing the automotive supply chain to business continuity, legal and reputational risk”.

The researchers also note that the costs of these recalls are escalating. One reason for that is growing consolidation and specialization within various product categories; many manufacturers today rely on the same few suppliers. So when a component or system used by multiple manufacturers is found to be defective, the financial impacts can reverberate widely.

Who is liable?

The researchers cited above also noted that the “risk implications (for autonomous technologies) are under-appreciated by large segments of the supply chain”.

I believe there are a couple of reasons for that. One is the pioneering nature of the technology. Although around 21 million self-driving cars are estimated to be on the road by 2026, mobile autonomy is in many ways still in its infancy. It was less than 15 years ago in 2005, for instance, that a driverless car named Stanley was the first autonomous vehicle to complete a 132-mile course in the Mojave Desert.

That means we are only just starting to collect data on how, and how often, different components and sub-components fail and what the implications of these failures are. Hence regulators around the world are insisting that autonomous vehicles undergo rigorous testing in closed environments before they are permitted on public roads.

Perhaps most important, many as-yet unanswered questions remain around the topic of apportioning liability when something goes wrong, currently a tangle of unresolved issues. Who will be liable, for instance, when a software error compromises performance? Or a manufacturing defect causes a sensor to malfunction? Or hackers trick an autonomous vehicle into veering toward oncoming traffic? (Researchers at a security firm recently demonstrated how such a lane-recognition hack could be accomplished simply by placing a set of small stickers along the roadway.)

In these and myriad other plausible scenarios, is it pure product liability or is the owner/operator of the vehicle also at least partially liable? And even if the issue is clearly product liability, who is at fault? The vehicle owner? The vehicle manufacturer? The firm(s) that designed the hardware? The OEM(s) that manufactured it? The software engineers who wrote the code?

A flexible, modular approach

AXA XL’s view is that the continued evolution of autonomous technologies shouldn’t depend on pre-emptively resolving all possible questions and risks.

So we’ve designed a multi-faceted insurance proposition that supports the design, development, testing and implementation of these technologies. The solution is aimed not only at companies that manufacture, own and operate autonomous vehicles but also the component makers and software developers that are part of this ecosystem.

A core element of the offering is a risk assessment based on a new benchmarking capability we created. The tool draws on historical data from the automotive and tech industries, enabling us to rank a client’s risks relative to the experiences of leading companies in these sectors. The insights derived from this assessment also can drive in-depth discussions about opportunities for minimizing the client’s exposures.

The proposition applies a modular approach to risk transfer; clients can structure various insurance coverages depending on their particular needs and circumstances. For a start-up running trials in a closed environment, for instance, the programme could include a traditional casualty policy along with some first-party property components. For more advanced companies operating autonomous vehicles in public spaces, the coverages can be scaled up to also include, for example, third-party motor liability, product recalls, malicious product tampering, cyber and on up to crisis management.

A final thought: Given the as-yet unanswered questions regarding the apportionment of liability, our view is that the contractual agreements between manufacturers and suppliers should delineate the risks each is prepared to assume, as well as the obligations of other participants in the application. In particular, whether or not a waiver-of-subrogation clause is included in these agreements will have a material impact on our underwriting and pricing decisions.

At AXA XL, we believe a collaborative approach is essential when working with autonomous technology clients and are already partnering with leading autonomous vehicle developers. They are the experts on their products, technologies, algorithms and so on. And we have expertise accumulated across multiple industries that can support them in making it to the next stage. By working together, we aim to help unleash the many benefits mobile autonomy offers while limiting the risks.

Mukadder Erdoenmez
Head International Casualty, Europe, AXA XL